[dpdk-dev] [PATCH] vhost: postpone ring addresses translations at kick time only

[dpdk-dev] [PATCH] vhost: postpone ring addresses translations at kick time only

[Maxime Coquelin]

If multiple queue pairs are created but all are not used, the
device is never started, as unused queues aren't enabled and
their ring addresses aren't translated. The device is changed
to running state when all rings addresses are translated.

This patch fixes this by postponning rings addresses translation
at kick time unconditionnaly, VHOST_USER_F_PROTOCOL_FEATURES
being negotiated or not.

Reported-by: Lei Yao <lei.a.yao@intel.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
 lib/librte_vhost/vhost_user.c | 33 ++++++++-------------------------
 1 file changed, 8 insertions(+), 25 deletions(-)

diff --git a/lib/librte_vhost/vhost_user.c b/lib/librte_vhost/vhost_user.c
index 1f6cba4b9..849755093 100644
--- a/lib/librte_vhost/vhost_user.c
+++ b/lib/librte_vhost/vhost_user.c
@@ -755,15 +755,12 @@ vhost_user_set_vring_kick(struct virtio_net **pdev, struct VhostUserMsg *pmsg)
 	RTE_LOG(INFO, VHOST_CONFIG,
 		"vring kick idx:%d file:%d\n", file.index, file.fd);
 
-	/*
-	 * Interpret ring addresses only when ring is started and enabled.
-	 * This is now if protocol features aren't supported.
-	 */
-	if (!(dev->features & (1ULL << VHOST_USER_F_PROTOCOL_FEATURES))) {
-		*pdev = dev = translate_ring_addresses(dev, file.index);
-		if (!dev)
-			return;
-	}
+	/* Interpret ring addresses only when ring is started. */
+	dev = translate_ring_addresses(dev, file.index);
+	if (!dev)
+		return;
+
+	*pdev = dev;
 
 	vq = dev->virtqueue[file.index];
 
@@ -845,29 +842,15 @@ vhost_user_get_vring_base(struct virtio_net *dev,
  * enable the virtio queue pair.
  */
 static int
-vhost_user_set_vring_enable(struct virtio_net **pdev,
+vhost_user_set_vring_enable(struct virtio_net *dev,
 			    VhostUserMsg *msg)
 {
-	struct virtio_net *dev = *pdev;
 	int enable = (int)msg->payload.state.num;
 
 	RTE_LOG(INFO, VHOST_CONFIG,
 		"set queue enable: %d to qp idx: %d\n",
 		enable, msg->payload.state.index);
 
-	/*
-	 * Interpret ring addresses only when ring is started and enabled.
-	 * This is now if protocol features are supported.
-	 */
-	if (enable && (dev->features &
-				(1ULL << VHOST_USER_F_PROTOCOL_FEATURES))) {
-		dev = translate_ring_addresses(dev, msg->payload.state.index);
-		if (!dev)
-			return -1;
-
-		*pdev = dev;
-	}
-
 	if (dev->notify_ops->vring_state_changed)
 		dev->notify_ops->vring_state_changed(dev->vid,
 				msg->payload.state.index, enable);
@@ -1315,7 +1298,7 @@ vhost_user_msg_handler(int vid, int fd)
 		break;
 
 	case VHOST_USER_SET_VRING_ENABLE:
-		vhost_user_set_vring_enable(&dev, &msg);
+		vhost_user_set_vring_enable(dev, &msg);
 		break;
 	case VHOST_USER_SEND_RARP:
 		vhost_user_send_rarp(dev, &msg);
-- 
2.13.6

Re: [dpdk-dev] [PATCH] vhost: postpone ring addresses translations at kick time only

[Maxime Coquelin]

Hi Lei,

On 11/03/2017 04:52 PM, Maxime Coquelin wrote:
> If multiple queue pairs are created but all are not used, the
> device is never started, as unused queues aren't enabled and
> their ring addresses aren't translated. The device is changed
> to running state when all rings addresses are translated.
> 
> This patch fixes this by postponning rings addresses translation
> at kick time unconditionnaly, VHOST_USER_F_PROTOCOL_FEATURES
> being negotiated or not.
> 
> Reported-by: Lei Yao<lei.a.yao@intel.com>
> Signed-off-by: Maxime Coquelin<maxime.coquelin@redhat.com>
> ---
>   lib/librte_vhost/vhost_user.c | 33 ++++++++-------------------------
>   1 file changed, 8 insertions(+), 25 deletions(-)

Could you confirm the patch fixes the issue on your side?

I tested below cases with and without IOMMU:
- Host DPDK queues = 1 / QEMU queues = 1 / Guest DPDK queues = 1
- Host DPDK queues = 2 / QEMU queues = 2 / Guest DPDK queues = 1
- Host DPDK queues = 2 / QEMU queues = 2 / Guest DPDK queues = 2

Thanks,
Maxime

Re: [dpdk-dev] [PATCH] vhost: postpone ring addresses translations at kick time only

[Yao, Lei A]

Hi, Maxime

> -----Original Message-----
> From: Maxime Coquelin [mailto:maxime.coquelin@redhat.com]
> Sent: Friday, November 3, 2017 11:57 PM
> To: dev@dpdk.org; yliu@fridaylinux.org; Yao, Lei A <lei.a.yao@intel.com>
> Cc: mst@redhat.com
> Subject: Re: [PATCH] vhost: postpone ring addresses translations at kick time
> only
> 
> Hi Lei,
> 
> On 11/03/2017 04:52 PM, Maxime Coquelin wrote:
> > If multiple queue pairs are created but all are not used, the
> > device is never started, as unused queues aren't enabled and
> > their ring addresses aren't translated. The device is changed
> > to running state when all rings addresses are translated.
> >
> > This patch fixes this by postponning rings addresses translation
> > at kick time unconditionnaly, VHOST_USER_F_PROTOCOL_FEATURES
> > being negotiated or not.
> >
> > Reported-by: Lei Yao<lei.a.yao@intel.com>
> > Signed-off-by: Maxime Coquelin<maxime.coquelin@redhat.com>
> > ---
> >   lib/librte_vhost/vhost_user.c | 33 ++++++++-------------------------
> >   1 file changed, 8 insertions(+), 25 deletions(-)
> 
> Could you confirm the patch fixes the issue on your side?
> 
> I tested below cases with and without IOMMU:
> - Host DPDK queues = 1 / QEMU queues = 1 / Guest DPDK queues = 1
> - Host DPDK queues = 2 / QEMU queues = 2 / Guest DPDK queues = 1
> - Host DPDK queues = 2 / QEMU queues = 2 / Guest DPDK queues = 2
> 
> Thanks,
> Maxime


Thanks for your patch. I test my test cases with your patch based on v17.11-rc2,
It can fix my issue here. 

BRs
Lei

Re: [dpdk-dev] [PATCH] vhost: postpone ring addresses translations at kick time only

[Maxime Coquelin]

On 11/06/2017 04:26 AM, Yao, Lei A wrote:
> Hi, Maxime
> 
>> -----Original Message-----
>> From: Maxime Coquelin [mailto:maxime.coquelin@redhat.com]
>> Sent: Friday, November 3, 2017 11:57 PM
>> To: dev@dpdk.org; yliu@fridaylinux.org; Yao, Lei A <lei.a.yao@intel.com>
>> Cc: mst@redhat.com
>> Subject: Re: [PATCH] vhost: postpone ring addresses translations at kick time
>> only
>>
>> Hi Lei,
>>
>> On 11/03/2017 04:52 PM, Maxime Coquelin wrote:
>>> If multiple queue pairs are created but all are not used, the
>>> device is never started, as unused queues aren't enabled and
>>> their ring addresses aren't translated. The device is changed
>>> to running state when all rings addresses are translated.
>>>
>>> This patch fixes this by postponning rings addresses translation
>>> at kick time unconditionnaly, VHOST_USER_F_PROTOCOL_FEATURES
>>> being negotiated or not.
>>>
>>> Reported-by: Lei Yao<lei.a.yao@intel.com>
>>> Signed-off-by: Maxime Coquelin<maxime.coquelin@redhat.com>
>>> ---
>>>    lib/librte_vhost/vhost_user.c | 33 ++++++++-------------------------
>>>    1 file changed, 8 insertions(+), 25 deletions(-)
>>
>> Could you confirm the patch fixes the issue on your side?
>>
>> I tested below cases with and without IOMMU:
>> - Host DPDK queues = 1 / QEMU queues = 1 / Guest DPDK queues = 1
>> - Host DPDK queues = 2 / QEMU queues = 2 / Guest DPDK queues = 1
>> - Host DPDK queues = 2 / QEMU queues = 2 / Guest DPDK queues = 2
>>
>> Thanks,
>> Maxime
> 
> 
> Thanks for your patch. I test my test cases with your patch based on v17.11-rc2,
> It can fix my issue here.

Thanks for the testing!
Maxime

> BRs
> Lei
> 

Re: [dpdk-dev] [PATCH] vhost: postpone ring addresses translations at kick time only

[Yuanhan Liu]

On Fri, Nov 03, 2017 at 04:52:35PM +0100, Maxime Coquelin wrote:
> If multiple queue pairs are created but all are not used, the
> device is never started, as unused queues aren't enabled and
> their ring addresses aren't translated. The device is changed
> to running state when all rings addresses are translated.
> 
> This patch fixes this by postponning rings addresses translation
> at kick time unconditionnaly, VHOST_USER_F_PROTOCOL_FEATURES
> being negotiated or not.
> 
> Reported-by: Lei Yao <lei.a.yao@intel.com>
> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

Acked-by: Yuanhan Liu <yliu@fridaylinux.org>

	--yliu

Re: [dpdk-dev] [PATCH] vhost: postpone ring addresses translations at kick time only

[Thomas Monjalon]

06/11/2017 12:58, Yuanhan Liu:
> On Fri, Nov 03, 2017 at 04:52:35PM +0100, Maxime Coquelin wrote:
> > If multiple queue pairs are created but all are not used, the
> > device is never started, as unused queues aren't enabled and
> > their ring addresses aren't translated. The device is changed
> > to running state when all rings addresses are translated.
> > 
> > This patch fixes this by postponning rings addresses translation
> > at kick time unconditionnaly, VHOST_USER_F_PROTOCOL_FEATURES
> > being negotiated or not.
> > 
> > Reported-by: Lei Yao <lei.a.yao@intel.com>
> > Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
> 
> Acked-by: Yuanhan Liu <yliu@fridaylinux.org>

Applied, thanks